Spatial-Similarity Dynamic Graph Bidirectional Double-Cell Network for Traffic Flow Prediction

Spatial-Similarity Dynamic Graph Bidirectional Double-Cell Network for Traffic Flow Prediction

Accurate traffic flow prediction plays a pivotal role in optimizing urban transportation systems and improving traffic management efficacy. To address the limitations of existing methods in modeling complex spatial-temporal dependencies within dynamic traffic networks, this paper introduces a Spatia...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhifei Yang, Zeyang Li, Jia Zhang
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Deep neural networks
traffic flow prediction
LSTM
GRU
Online Access:https://ieeexplore.ieee.org/document/11031461/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Accurate traffic flow prediction plays a pivotal role in optimizing urban transportation systems and improving traffic management efficacy. To address the limitations of existing methods in modeling complex spatial-temporal dependencies within dynamic traffic networks, this paper introduces a Spatial-Similarity Dynamic Graph Bidirectional Double-Cell Network (SDGBDCN). The proposed architecture incorporates two innovative components: 1) a Spatial Similarity Dynamic Graph Convolution (SDGCN) module that adaptively aggregates spatial features through node similarity analysis and time-varying graph structures, and 2) a Bidirectional Double-Cell Recurrent Neural Network (Bi-DouCRNN) that combines LSTM and GRU mechanisms via dual-gating operations to capture multi-scale temporal dynamics. Comprehensive evaluations on PeMS datasets demonstrate superior performance compared to existing approaches. Statistical validation through AIC and SBIC metrics confirms the model’s exceptional capability, achieving record-low scores of 1204.2 and 1219.8 respectively. This research advances traffic prediction methodologies through its integrated approach to dynamic spatial correlation modeling and bidirectional temporal learning, providing valuable insights for intelligent transportation system development.
ISSN:2169-3536

Similar Items